gwatch-cu128 0.0.3

G-Watch is a toolbox for GPU profiling and program analysis.

G-Watch

G-Watch is a toolbox for GPU profiling and program analysis. It features rich Profiling capabilities on both NVIDIA and AMD GPUs. Additionally, G-Watch offers Program Analysis tools for inspecting compiler-generated GPU binaries, facilitating secondary development tasks such as register analysis and binary instrumentation.

Features

We would release guidance for using G-Watch soon, currently please take refer to examples

Installation

Install from package

You can install G-Watch directly from PyPI:

# prerequisites for gwatch
apt-get update
apt-get install -y  python3 python3-pip libelf-dev protobuf-compiler libwebsockets-dev libnuma-dev

# install from PyPI source
pip install gwatch==0.0.2

Install from source [CUDA]

1. Clone this repository

   git clone --recursive https://github.com/G-Watch/G-Watch-dev.git
   

2. Start official nvidia container for building and testing

   cd scripts/docker
   bash run_cuda_12_8.sh -s 1  # start and enter a container with id 1
   
   # don't need to run, just noted here
   bash run_cuda_12_8.sh -e 1  # enter a container with id 1
   bash run_cuda_12_8.sh -c 1  # close and remove a container with id 1
   

3. Install prerequisites

   # inside the container
   apt-get update
   apt-get install -y  git pkg-config python3 python3-pip cmake meson build-essential libeigen3-dev \
                       python3-dev wget libelf-dev libwebsockets-dev libnuma-dev protobuf-compiler \
                       gdb curl libcurl4-openssl-dev xxd libdwarf-dev python-is-python3
   pip3 install perfetto PyYAML tqdm pyyaml packaging loguru PrettyTable matplotlib pandas ninja==1.11.1.3 --break-system-packages
   
   # for most of our test workloads, we use PyTorch, so it's recommended to install torch as well
   # https://pytorch.org/get-started/locally/
   pip3 install torch torchvision torchaudio --break-system-packages
   

4. Build and install python package from source

   # inside the container
   cd /root
   
   python3 setup.py clean  
   
   # build and install gwatch
   python3 setup.py install
   

Install from source [ROCm]

1. Clone this repository

   git clone --recursive https://github.com/G-Watch/G-Watch-dev.git
   

2. Start official nvidia container for building and testing

   cd scripts/docker
   bash run_rocm_7_2.sh -s 1  # start and enter a container with id 1
   
   # don't need to run, just noted here
   bash run_rocm_7_2.sh -e 1  # enter a container with id 1
   bash run_rocm_7_2.sh -c 1  # close and remove a container with id 1
   

3. Install prerequisites

   # inside the container
   apt-get update
   apt-get install -y  git pkg-config python3 python3-pip cmake meson build-essential libeigen3-dev \
                       python3-dev wget libelf-dev libwebsockets-dev libnuma-dev protobuf-compiler \
                       gdb curl libcurl4-openssl-dev xxd libdwarf-dev python-is-python3 libdw-dev libsqlite3-dev
   pip3 install perfetto PyYAML tqdm pyyaml packaging loguru PrettyTable matplotlib pandas ninja==1.11.1.3 --break-system-packages
   
   # for most of our test workloads, we use PyTorch, so it's recommended to install torch as well
   # https://pytorch.org/get-started/locally/
   pip3 install torch torchvision --index-url https://download.pytorch.org/whl/rocm7.1
   

4. Build and install python package from source

   # inside the container
   cd /root
   
   python3 setup.py clean  
   
   # build and install gwatch
   python3 setup.py install